A Systematic Comparison Between Classical Optimal Scaling and the Two-Parameter IRT Model
Matthijs J. Warrens, Dato N. M. de Gruijter, Willem J. Heiser · Applied Psychological Measurement · 2007
In this article, the relationship between two alternative methods for the analysis of multivariate categorical data is systematically explored. It is shown that the person score of the first dimension of classical optimal scaling correlates strongly with the latent variable for the two-parameter item response theory (IRT) model. Next, under the assumption that the latent variable has a standard normal distribution, it is derived how the IRT item parameters are related to the optimal category weights.